#076 - Why 90% of Enterprise AI Pilots Fail - with Scribe CEO Jennifer Smith

This week on GAEA Talks Live from HumanX, Graeme Scott sits down with Jennifer Smith - co-founder and CEO of Scribe, former Head of Business Development and Partnerships at Greylock Partners, and one of the clearest thinkers in the world on enterprise AI transformation.Jennifer's journey is the perfect setup for what she is building today. After Princeton and an MBA from Harvard, she spent seven years at McKinsey in the operations practice, flying to operations centres across America. She watched first hand how the best operator on any team had their own personal best practice that bore no resemblance to what the rest of the team had been trained on. She then moved into venture capital at Greylock Partners, where she interviewed over twelve hundred CIOs and CTOs and asked them all the same question: what do you actually want Silicon Valley to build? The same answer kept coming back. Institutional knowledge. That insight became Scribe, the context layer for workflow data across the enterprise. The platform now works with nearly half of the Fortune 500 and over ninety thousand customer organisations.In this episode, recorded live at HumanX 2026 in San Francisco, Jennifer makes the case that the reason ninety percent of enterprise AI pilots fail is not the model. It is the missing context layer underneath. She walks Graeme through why dropping the most intelligent AI in the world into a business with no context is like hiring an army of Nobel laureates and refusing to onboard them. She lays out a complete reframe of how enterprise leaders should approach AI transformation: stop starting with strategy, start by mapping how your business actually works. She explains why she would not have been comfortable deploying agents at scale a year ago, why she is genuinely comfortable with it now, and what business leaders should do in the next twelve to eighteen months to make sure they are on the right side of the curve. This is one of the most useful conversations on practical enterprise AI that GAEA Talks has recorded.What you'll take away from this conversation:• The context layer thesis - why the missing piece in every failed enterprise AI pilot is not the model, but the data on how the business actually works• The Nobel laureate analogy - why dropping the world's smartest intelligence into a business with no onboarding is exactly why ROI keeps disappearing• The Stanford and MIT studies on why ninety percent plus of enterprise AI pilots fail• Why most enterprises are doing AI transformation in the wrong order• The "before map and after map" framework for measuring real AI ROI• Why engineering and customer support are the only two functions where AI is delivering at scale today• The Waymo lesson - map a process for "millions of miles" with humans before handing it to an agent• Why so many founders in San Francisco are quietly anxious they are already behind• The mandate versus strategy distinction - "do more with less using AI" is not a strategy• Why the limiting factor in AI adoption is no longer the technology, it is the human and organisational ability to absorb change• The shift from one human equals one output to systems of work that can be filled by humans or agents• The two-week Scribe pilot that maps workflows and builds AI transformation business cases• Why the future of work will be a mix of humans and agents - and why leaders must make the work legible to both• Jennifer's advice for senior leaders: define success first, then map operations, then build ROI-backed cases